摘要
重采样技术是改进标准粒子滤波器粒子退化的关键技术,但也造成了样本枯竭。针对重采样技术,提出了一种基于社会蜘蛛群优化的粒子滤波算法。针对样本枯竭问题,一方面依据有效粒子数,设置阈值判断是否重采样;另一方面依据可复制粒子数,以高斯分布在大权重粒子均值附近随机选取粒子,保证了算法的实时性以及粒子多样性。针对算法的收敛能力,首先依据群体突变思想,当迭代次数达到设置的动态次数因子,初始化粒子;其次依据粒子的寻优速度和程度因子,动态自适应调整粒子权重;最后依据遗传算法的交叉概率,限制新粒子诞生速度,从而提高算法的整体收敛性能。通过对比可知,该算法的整体性能优于其他改进算法,能够有效解决非线性误差对滤波精度的影响。
Resampling technology is a key technology to improve the particle degradation of standard particle filters,but it also causes sample depletion.Aiming at the resampling technique,a particle filter algorithm based on social spider optimization is proposed.To solve the problem of sample depletion,on one hand,a threshold is set to determine whether to resample or not based on the number of effective particles.On the other hand,particles are randomly selected with Gaussian distribution near the mean value of large-weight particles based on the number of replicable particles,ensuring the real-time performance of the algorithm and the diversity of particles.Aiming at the convergence ability of the algorithm,firstly,according to the idea of group mutation,when the number of iterations reaches the set dynamic number factor,the particles will be initialized.Secondly,the particle weight will be adjusted dynamically and adaptively according to the optimization speed and degree factor of the particles.Finally according to the crossover probability of genetic algorithm the birth speed of new particles is limited,thereby improving the overall convergence performance of the algorithm.According to comparison,the overall performance of this algorithm is better than that of other improved algorithms,and the algorithm can effectively solve the effect of nonlinear errors on the filtering accuracy.
作者
谷旭平
唐大全
GU Xuping;TANG Daquan(School of Aviation Operations and Support,Naval Aviation University,Yantai 264001,China)
出处
《自动化仪表》
CAS
2021年第8期37-43,49,共8页
Process Automation Instrumentation
关键词
粒子滤波
社会蜘蛛群
重采样
样本枯竭
群体突变
交叉概率
自适应权重
粒子多样性
Particle filter
Social spider swarm
Resampling
Sample depletion
Group mutation
Crossover probability
Self-adaptive weight
Particle diversity
